Why this matters
Most small businesses answer the same customer questions every week: hours, prices, service area, booking and refunds. The answers are already in the FAQ, yet someone still types them, and anyone who writes after hours waits until morning.
Chatbase builds an AI agent trained on your own content and puts it on your website as a chat bubble. Told to answer only from your FAQ, it answers the routine questions straight away and points the rest to a person.
Practice
Do every step yourself in a practice copy of the real tool. It checks your work, and “Do it for me” finishes any step you get stuck on.
What you’ll build
You help Dana Ruiz, who runs Juniper Lane Bookkeeping, a made-up five-person firm that answers about 40 repeat questions a week by email.
You create a Chatbase agent, train it on a sample FAQ file, write General instructions and Guardrails with an "I'm not sure" fallback, keep Temperature low, compare versions, test real questions and one it should not know, look at the Escalations action, find the Website widget code and lock it to your domain, and fix an answer from the chat logs. The Chatbase screens are screenshots from Chatbase's own docs, credited and linked under each one.
What you need
- Nothing to start the practiceThe lesson uses Chatbase's own docs screenshots. Juniper Lane Bookkeeping and every number are made up.
- The sample FAQ22 made-up questions and answers as a .txt file, ready to upload to Chatbase.
- A Chatbase account, to do it for realFrom chatbase.co/pricing, read 18 October 2026: Hobby is $40 a month ($32 billed yearly) with 700 message credits, Standard is $150 a month ($120 billed yearly), Pro is $500 a month ($400 billed yearly), and the page offers a 7-day trial. It lists no free plan.
- Your own FAQUp to date, with selectable text if it is a PDF. The agent is only as right as this file.
Concepts
Train it on your own answers
On Data sources you add files, website pages, text snippets, Q&A pairs, Notion pages or tickets. Chatbase accepts .pdf, .txt, .doc and .docx files up to 20 MB, and each source shows Syncing, then Synced.
When a price or policy changes, update the source. The agent says what its sources say.
The fallback is your hand-off
In Build > Instructions, General sets the business, role and tone. Guardrails are rules for every chat. Chatbase's Best Practices page recommends telling the agent to answer only from its training data, with a fallback answer for anything else. Put your email and phone number in that fallback.
For tickets, the Escalations action creates one in a helpdesk you have already set up.
Test first, then read the logs every week
Ask real customer questions in the Playground, plus one it should not know, before you install the embed code. After it goes live, Activity shows every chat, and Revise answer turns a fix into a new Q&A source.
Do it for real
Now repeat it in your own account. Tick each step as you go.
Quick check
3 questions to check you’ve got it.
1. A customer asks something your FAQ does not cover. What should the agent do?
- Give its best guess
- Say it is not sure and point to a person
- Ignore the question
Show the answer
Say it is not sure and point to a person. A Guardrails rule with a fallback answer stops the agent guessing prices or policies, and hands the customer to you.
2. Which files can you upload as a Chatbase source?
- Only websites
- .pdf, .txt, .doc and .docx up to 20 MB
- Images of your price list
Show the answer
.pdf, .txt, .doc and .docx up to 20 MB. Chatbase's docs list those four file types, and a PDF needs selectable text.
3. The agent gave a wrong answer last week. What is the quickest fix?
- Delete the agent and start again
- Open the chat in Activity and use Revise answer
- Turn the temperature up
Show the answer
Open the chat in Activity and use Revise answer. Revise answer saves the right answer as a new Q&A source, so the agent uses it next time.
KDAN: Get a Vendor Contract's Key Terms with AI Chatbot
You built an agent that answers your FAQ. Next in this course: KDAN: Get a Vendor Contract's Key Terms with AI Chatbot.
Start module 34 →